When you need to get data out of an image, usually barcodes, fiducials or computer vision markers are a good idea. Classic 2D barcodes allow decoding of a number (e.g. a product number on a product), fiducials mark very precisely a specific point (if you want to align a machine with something on a flat conveyor belt) and computer vision markers do both of these in arbitrary dimensions. They allow encoding information (usually an id number) and carry known geometrical information that allow computing a camera position in relation to the marker. On top of this they should be easy to find in an camera image, should be somewhat robust against bad imaging conditions and occlusion and should be as precise as possible. That’s a lot of requirements that come with different kinds of tradeoffs. Not very surprisingly, there are a lot of different markers.
So, I did what every rational person would do and added yet another one: (from the abstract)
We present Seedmarkers, shape-independent topological markers that can be embedded in physical objects manufactured with common rapid-prototyping techniques. Many markers are optimized for technical performance while visual appearance or the feasibility of permanently merging marker and physical object is not considered. We give an overview of the aesthetic properties of a wide range of existing markers and conducted a short online survey to assess the perception of popular marker designs. Based on our findings we introduce our generation algorithm making use of weighted Voronoi diagrams for topological optimization. With our generator, Seedmarkers can be created from technical drawings during the design process to fill arbitrary shapes on any surface. Given dimensions and manufacturing constraints, different configurations for 3 or 6 degrees of freedom tracking are possible. We propose a set of application examples for shape-independent markers, including 3D printed tangibles, laser cut plates and functional markers on printed circuit boards.
The fancy thing about Seedmarkers is that they can be generated for (almost) any shape on a flat surface and integrated into it during manufacturing. This is especially handy if you 3d-print something or cut it with a lasercutter.